A new approach to managing temperature processes

Researchers from the Belgrade Faculty of electrical engineering and Maribor Faculty for electrical engineering and computer sciences have joined their research experiences in order to create a new approach to managing temperature processes.
NEURO-TEMP is an acronym for a project whose full name is “New approach to managing temperature processes based on “soft-computing””, which is financed by EUREKA program of the European Union. EUREKA is a program which stimulates and promotes cooperation between enterprises and institutions in the field of advanced technologies. Through that, the idea for NEURO-TEMP’s proposition came to pass: researchers from two faculties (Faculty of electrical engineering, Belgrade, and Faculty of electrical engineering and computer science, Maribor) and two research centers (WBC from Belgrade and The Rogaška Slatina Network) from two countries, Serbia and Slovenia, have decided to combine their research efforts and propose a new approach to managing temperature processes.

The idea behind this research originated from the experiences of these institutions’ researchers in solving different problems in their industry, especially in manufacturing and exploitation of industrial furnaces. The problems characteristic for this particular industry are extremely complex, mainly because of the distributed nature of temperature fields and also because of the non-linear and non-stationary nature of the processes. For example, industrial furnaces are often rather large, since they serve for thermal processing of parts for ship engines, propellers, bulldozer spoons and such. It is extremely important that an adequate temperature profile in the interior of the furnace is achieved during the thermal process, while that process is simultaneously homogenous in spatial distribution. Unfortunately, the control systems which manage the heaters and ventilators have only a limited number of measuring points, and based on that sum of limited measurements they generate controlling signals for the corresponding actuators. The most often used laws of management are PI and PID regulators; however, they are not the natural solutions for this type of management objects.

The NEURO-TEMP project displays and brings to fruition the idea of basing the new management solution on the models of processes which have originated from the method of finite elements, where corresponding neural networks are trained with the goal of numeric implementation, while the laws of management are rooted in the spirit of logic. The method of finite elements enables solving of partial differential equations, which are the natural way of describing the temperature processes taking place in industrial furnaces. Having in mind their numeric complexity, they are not applicable in real-time and cannot be part of the management loop. The idea embodied in the project is to train artificial neural networks which are to simulate the solution to the method of finite elements, and to be included in feedback during the temperature process management in a numerically simpler manner.

It is important to point out that some of these algorithms have already been installed in real industrial facilities, such as Bosio LLC in Slovenia, where these types of management solutions are still in the test phase. It is shown that this approach to managing heaters and ventilators during thermal processing of material provides significantly more homogenous processing, which consequentially results in better product quality and a longer life cycle. On the other hand, the nature of projected algorithms also enables direct diagnostics of certain sensors and actuators, which means a significant step forward in terms of equipment availability, environmental safety and overall system reliability.

Source: www.innovationfund.rs

Oct 10, 2012 in Serbia
tag: , ,